PrologMCP: A Standardized Prolog Tool Interface for LLM Agents

2026-06-12

AI Summary

Large language models exhibit limited performance on deep deductive reasoning tasks, and their internal chain-of-thought reasoning incurs substantial computational costs. To address this, this work proposes a symbolic delegation mechanism that offloads logical reasoning to an external Prolog solver via a standardized interface. The authors introduce the first universal, open-source, stateful Prolog tool server built upon the Model Context Protocol (MCP), enabling an iterative translate-execute-check-repair reasoning loop. By integrating session isolation and structured error reporting, the approach achieves near-perfect accuracy on the PARARULE-Plus benchmark—scoring 1.00 on the general subset and 0.99 on the challenge subset—significantly outperforming both standard and reasoning-augmented large language models.

Abstract

Frontier reasoning-tuned language models still fail on deductive tasks at depth, and the cost of improved performance through extended internal reasoning scales poorly. Symbolic delegation offers a complementary route: a language model translates the problem, while a solver performs the inference. However, current autoformalization pipelines for logic programming are typically bespoke integrations tied to particular tasks or agents. We introduce PrologMCP, a task-agnostic, open-source server that exposes Prolog as a stateful tool through the Model Context Protocol (MCP). Its compact tool interface, structured error reporting, and per-session isolation make the translate-run-inspect-repair loop a reusable primitive for MCP-capable agents. We evaluate a formalizer agent enhanced with PrologMCP against standard and reasoning LLMs (Claude Sonnet 4.6, GPT-4.1, and o4-mini) on two subsets of PARARULE-Plus: a general-purpose sample and a more challenging one targeting a specific failure mode of natural-language reasoning. On the general sample, the formalizer matches or exceeds reasoning LLMs (accuracy 1.00 vs.\ 1.00 / 0.998), with the largest gains over standard models (0.762 for GPT-4.1). On the challenging subset, the formalizer remains near-perfect (1.00 / 0.99) while reasoning LLMs drop to 0.95 / 0.94. These results suggest that delegating inference to Prolog via MCP is a robust and inspectable alternative to extended natural-language reasoning.

Problem

• deductive reasoning
• autoformalization
• logic programming
• LLM agents
• symbolic delegation

Innovation

• PrologMCP
• symbolic delegation
• Model Context Protocol
• autoformalization
• deductive reasoning